whisperText

AUDIOTEXT
ready
no signal

Drop in a recording — an interview, a meeting, a voice memo — and get text with timestamps, ready to download as TXT, SRT, VTT, DOCX or PDF. Everything is computed on your computer: the file never leaves your browser and is never sent to any server.

How it works
  1. The browser decodes the file and converts the audio to 16 kHz mono — what the model expects.
  2. The recording is split into five-minute segments, with boundaries placed in pauses so a cut never lands mid-word.
  3. Whisper runs the transcription on your graphics card via WebGPU, or on the CPU when there's no GPU. Slower, but it always works.
  4. Text appears as it's produced. The timestamp next to each paragraph jumps the player to exactly that spot.

The first run downloads the model from Hugging Face servers (200 MB – 1.5 GB, depending on your choice). After that it stays in the browser's storage and the app works offline.

Formats: anything your browser can play — MP3, WAV, M4A, OGG, FLAC, WEBM, as well as audio tracks of video files.

Drop an audio file or click to choosemp3, wav, m4a, ogg, flac, webm — the file is never uploaded
Whispertext — Audio → Text